The the intersection between photography and artificial intelligence has become one of Google’s main interests, and from what little we know so far, RealFill is the next step in that direction, and a really interesting and promising one at that. And even before 2022, but especially since 2022, computational photography is emerging as one of the key points by which smartphone manufacturers and providers of photo and video-related services try to differentiate themselves from their competitors.

This is a competition in which Google has been able to place itself very well., and which subsequently turned the Pixel smartphones into one of the most reliable proposals on the market in this sense, even surpassing Apple in some aspects, which, as we remember, was the predecessor of the focus (or one of them) on the advantages of mobile phones. phones in their photography and video recording capabilities. Competition that also benefits users, which translates into continuous improvement.

In the case of Googlewe’ve already seen some AI-based computational photography features which have become a great success, such as the magic eraser and the magic editor, both recently released by Google for all interested users, or the ability to individually select in a series the facial shot that one of the people playing in it came out the best. For group photos, from three to four people, the usefulness of this function is exceptional.

PICTURE-

In this context, as we read in IT Home (in Chinese) Google has just registered a ReadFill feature that allows you to fill images with patterns to create generative patterns, i.e. through artificial intelligence. In this way, as we can see in the samples published by the mentioned medium, we can both “correct” those on which some part is not well displayed, and complete those that we want to enlarge relative to the original frame.

ReadFill Based on a study conducted by Google and Cornell University which, entitled “Reference-based generation of real images”, proposed the necessary techniques for a generative model to be able to complete an image from subsamples extracted from other shots. This of course greatly improves other AI-based “fill” suggestions that are based solely on the content of the image to be enlarged.

At the moment, we don’t know when this technology will be introduced, but the fact that it was registered just over two weeks before the Google I/O 2024 celebration encourages us to be optimistic, and therefore to think that We won’t have to wait long to see it in action..